1. Expanding Mobile Wireless Capacity -s2.0-S0308596113001900-Main

8/12/2019 1. Expanding Mobile Wireless Capacity -s2.0-S0308596113001900-Main

1/16

Expanding mobile wireless capacity: The challenges presentedby technology and economics$,$$

Richard N. Clarke n

AT&T, Suite 400 North, 1120 20th Street NW, Washington, DC 20036, USA

a r t i c l e i n f o

Keywords:

Spectrum

Reuse

Mobile

Wireless

Network

LTE

a b s t r a c t

As demand for mobile broadband services continues to explode, mobile wireless networks

must expand greatly their capacities. This paper describes and quantifies the economic

and technical challenges associated with deepening wireless networks to meet this

growing demand. Methods of capacity expansion divide into three general categories:

the deployment of more radio spectrum; more intensive geographic reuse of spectrum;

and increasing the throughput capacity of each MHz of spectrum within a given

geographic area. The paper describes these several basic methods to deepen mobile

wireless capacity. It goes on to measure the contribution of each of these methods to

historical capacity growth within U.S. networks. The paper then describes the capabilities

of 4G LTE wireless technology, and further innovations off of it, to further improve

network capacity. These capacity expansion capabilities of LTE-Advanced along with

traditional spectrum reuse are quantified and compared to forecasts of future demand to

evaluate the ability of U.S. networks to match future demand. Without significantly

increasing current spectrum allocations by 560 MHz over the 20142022 period, the

presented model suggests that U.S. wireless capacity expansion will be inadequate to

accommodate expected demand growth. This conclusion is in contrast to claims that the

U.S. faces no spectrum shortage.

& 2013 The Author. Published by Elsevier Ltd. All rights reserved.

1. Overview

Demand for mobile wireless services continues to explode. Cisco's latest Visual Networking Index (VNI) reports that

global mobile data traffic grew 70% in 2012,driven by average connection speeds that more than doubled from 248 kbps

to 526 kbps (Cisco, 2013, p. 1). Further, Cisco estimates that by 2017, global mobile data traffic will exceed its 2012 level by a

factor of 12.6. While in many parts of the world, significant portions of expansion in mobile wireless network capacity will

continue to be due to expansions in the geographic coverage of wireless data networks, in developed countries such as the

Contents lists available at ScienceDirect

URL: www.elsevier.com/locate/telpol

Telecommunications Policy

0308-5961/$- see front matter & 2013 The Author. Published by Elsevier Ltd. All rights reserved.http://dx.doi.org/10.1016/j.telpol.2013.11.006

$ This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution,

and reproduction in any medium, provided the original author and source are credited. The analyses and data presented in this paper are intended to portray the U.S. mobile wireless industry on a national average basis. They may not be

representative of any particular U.S. geographic region or mobile operator, including AT&T. No proprietary AT&T data were used in performing these

analyses. The conclusions developed in this paper are those of the author, alone, and should not be construed as representing any official position of AT&T. I

am indebted to my colleagues at AT&T and Peter Rysavy for valuable assistance in preparing this paper. Very useful suggestions were also received from

two anonymous reviewers. All remaining errors are my own. An earlier version of this paper was presented at the 19th Biennial Conference of the

International Telecommunications Society in Bangkok on November 20, 2012.n Tel.: 1 202 457 2130.

E-mail address: [email protected]

Telecommunications Policy ] (]]]]) ]]]]]]

Please cite this article as: Clarke, R. N. Expanding mobile wireless capacity: The challenges presented by technology andeconomics.Telecommunications Policy (2013), http://dx.doi.org/10.1016/j.telpol.2013.11.006i
http://www.sciencedirect.com/science/journal/03085961http://www.elsevier.com/locate/telpolhttp://dx.doi.org/10.1016/j.telpol.2013.11.006mailto:[email protected]://dx.doi.org/10.1016/j.telpol.2013.11.006http://dx.doi.org/10.1016/j.telpol.2013.11.006http://dx.doi.org/10.1016/j.telpol.2013.11.006http://dx.doi.org/10.1016/j.telpol.2013.11.006http://dx.doi.org/10.1016/j.telpol.2013.11.006http://dx.doi.org/10.1016/j.telpol.2013.11.006http://dx.doi.org/10.1016/j.telpol.2013.11.006http://dx.doi.org/10.1016/j.telpol.2013.11.006http://dx.doi.org/10.1016/j.telpol.2013.11.006http://dx.doi.org/10.1016/j.telpol.2013.11.006http://dx.doi.org/10.1016/j.telpol.2013.11.006http://dx.doi.org/10.1016/j.telpol.2013.11.006http://dx.doi.org/10.1016/j.telpol.2013.11.006http://dx.doi.org/10.1016/j.telpol.2013.11.006mailto:[email protected]://dx.doi.org/10.1016/j.telpol.2013.11.006http://dx.doi.org/10.1016/j.telpol.2013.11.006http://dx.doi.org/10.1016/j.telpol.2013.11.006http://www.elsevier.com/locate/telpolhttp://www.sciencedirect.com/science/journal/03085961


2/16

U.S., advanced mobile broadband networks already cover 98.5% of potential subscribers (FCC, 2011, 46). Thus, the network

expansions necessary to accommodate demand growth in developed countries will focus most greatly on deepening

network capacities. Technically and economically, this presents a different set of challenges from simply expanding coverage

scope a topic that has been addressed extensively in universal service research.1

The purpose of this paper is to describe and quantify the challenges particularly associated with wireless network

deepening. This includes an analysis of the technical issues concerning what techniques for capacity deepening are feasible,

and also consideration of the costs of these techniques to determine the economic capability of these techniques to keep up

with growing demand. While certain parties have suggested that improvements in technology to increase throughputcapacity per megahertz (MHz) of spectrum and increased geographic reuse of spectrum will be adequate to address wireless

demand growth in the U.S. over the next five to ten years, this analysis finds otherwise.2 Although methods to improve

throughput capacity per MHz or increase spectrum reuse may be technologically feasible, and are expected to be used

intensively by wireless providers, by themselves they are likely to be inadequate or to become uneconomic absent

significant increases in mobile wireless pricing. Thus, even under the conservative modeling assumptions of this paper,

substantial quantities of additional spectrum (on the order of 560 MHz) will need to be deployed for mobile wireless use if

currently forecasted demand is to be satisfied over the next decade without significant service quality degradation or

rationing from price increases.

This finding is consistent with the conclusions developed by several other studies in the literature that have examined

the ability of current or expected spectrum assignments and technologies to accommodate forecasted demand.3

The analysis presented in this study will differ from these prior efforts both by improving on the accuracy of their analyses

and by projecting certain enhancements in the ability of evolving wireless technologies to carry more mobile traffic.

This paper begins by describing the basic techniques that may be used to expand mobile wireless capacity. These includeincreasing raw amounts of available radio spectrum, increasing the absolute carrying capacity of each MHz of spectrum,

reducing the bandwidth required to carry popular applications, and increasing the utilization of each MHz of spectrum or

unit of infrastructure through cell-splitting, sharing or multiple use. InSection 3, this history of technological evolution is

contrasted with both the growth in available mobile wireless spectrum and the growth in mobile wireless demand.

The paper goes on to catalog the possible forward-going capabilities and economics of several of the most well-known

potential Fourth Generation (4G) Long Term Evolution (LTE) wireless technology innovations, including innovations whose

effects remain highly speculative. By comparing the joint capacity expansion capabilities of these new and old techniques

with demand growth estimates, it is possible to evaluate their ability to accommodate demand growth and to reduce

upwards pressure on current wireless pricing. In the end, the analysis demonstrates that by themselves, these methods will

be inadequate to accommodate fully expected demand growth at today's prices. Thus, increased assignment of radio

spectrum to mobile wireless will be essential. This is in contrast to suggestions from certain parties that spectrum scarcity

should not be a terribly significant concern for government policymakers.

2. Mobile wireless capacity expansion techniques

Methods for expanding mobile network capacity divide into three general categories: the deployment of more radio

spectrum; more intensive geographic reuse of spectrum; and increasing the throughput capacity of each MHz of spectrum

within a given geographic area.

The carrying capacity of a mobile wireless system is the total amount of data or voice traffic that the system is able to

transfer to and from customers.4 Wireless data are carried by modulating or distorting radio waves. The quantity of waves

(or amount of spectrum) a wireless system is allowed to modulate each second is called its bandwidth, and is measured in

hertz (Hz). Everything else equal, a signal with a higher bandwidth (i.e., more Hz) can carry more data per second than a

signal of lower bandwidth (i.e., less Hz).

The total amount of data that a network may transfer over a given period of time relates closely to the rate at which ittransfers data bytes. All things equal, a faster network will transfer more bytes than a slower network. Rates of data transfer

are measured in terms of bits per second (bps).5 Note, however, that in addition to raw transmission speed, the total amount

of data transfer will be higher on a network that operates as a higher usage/fill factor (i.e., transfers data during more

seconds of the measurement period). This can be achieved if a network has traffic offered more uniformly to it over the

measurement period either because the network serves multiple users whose patterns for offering traffic to the network

1 See, for example,Williams et al. (2011).2 SeeBazinet and Rollins (2011), Bode (2012), Burstein (2011), Chen (2012a,b)and Reardon (2010) for arguments suggesting that the U.S. faces no

serious shortage of mobile wireless spectrum.3 See FCC (2010b), Feldman et al. (2011), Lawson (2012), Rysavy (2011) and Rysavy Research (2010, 2011a,c) for analyses suggesting that mobile

wireless spectrum will become critically scarce.4 In general, the capacity of the total network will be limited by the capacity of its last-mile radio access network (RAN). Although congestion on fixed

backhaul links may possibly occur, increased availability of fiber backhaul facilities should allow backhaul bottlenecks to be engineered away.

5 By convention, data transfer rates are typically measured in terms of bits (b), and data quantities in terms of bytes (B). Because there are eight bits in abyte, a transfer rate of 8 bps corresponds to transferring one byte per second (1 Bps).

Please cite this article as: Clarke, R. N. Expanding mobile wireless capacity: The challenges presented by technology andeconomics. Telecommunications Policy (2013), http://dx.doi.org/10.1016/j.telpol.2013.11.006i

R.N. Clarke / Telecommunications Policy ] (]]]]) ]]]]]]2
http://dx.doi.org/10.1016/j.telpol.2013.11.006http://dx.doi.org/10.1016/j.telpol.2013.11.006http://dx.doi.org/10.1016/j.telpol.2013.11.006http://dx.doi.org/10.1016/j.telpol.2013.11.006http://dx.doi.org/10.1016/j.telpol.2013.11.006


3/16

are less than perfectly correlated, or because the network employs packet scheduling protocols that efficiently divide traffic

into different queues based on the immediacy of their need for transmission.6

Perhaps the most well-known way for cellular networks to increase the amount of data they carry is by dividing or

splitting cells to reduce cell size, and thus increase the number of cells serving a given area. This is done by deploying more

radio towers/antennas and shrinking the reach of each tower by reducing the radiated power of its radio transmissions. This

allows radio spectrum to be reusedfor multiple simultaneous transmissions within the geographic area. Thus by subdividing

cells, the amount of traffic that a Hz of spectrum can carry within an overall geographic area (measured by bps/km 2) is

increased. But while very effective at deepening wireless network capacity, this method is also expensive requiring theconstruction of extra towers/antennas, deploying more radios and base station equipment; as well as extending additional

backhaul links to link new towers back into the mobile operator's core network.7

Over the past thirty years, mobile wireless networks have evolved from First Generation (1G) to Fourth Generation (4G)

technologies that make increasingly efficient use of available radio spectrum. In the following sections, the contributions of

4G technical enhancements are developed and compared with contributions to capacity growth from raw spectrum

growth and increased reuse of available radio frequencies.

3. Analysis of historical demand and spectrum growth

3.1. Growth in mobile usage

The history of U.S. mobile services demand has been one of continuous and increasingly rapid growth. If voice minutes

are converted into data traffic equivalents, it is possible to chart historical rates of total mobile demand growth. This isshown inFig. 1. As can be seen, U.S. data usage outstripped voice usage by the end of 2009, and by 2012 data usage exceeded

voice usage by a factor of 7.5.8

This explosive growth in data demand has been the product of two forces: increasing numbers of subscribers adopting

advanced mobile data devices and increasing monthly use of mobile data services by the subscribers that are using these

devices. Because adoption of smartphones is now nearing 70% in the U.S., it is possible that the first force driving mobile

usage growth may begin to stabilize, but the latter shows no hint of flagging. 9 Cisco estimates that while current mobile data

growth rates may be the maximum that may be seen in the advanced countries, these growth rates will decline only

slowly.10 Total mobile traffic growth is shown in Fig. 2.11

3.2. Growth in spectrum resources

In contrast to the exponential growth in mobile bandwidth demand, U.S. mobile spectrum has been doled out at a far

more moderate pace. Table 1 illustrates the significant assignments of wireless spectrum by the Federal Communications

Commission (FCC) for U.S. mobile use.

WhileTable 1lists current mobile wireless spectrum allocations by the approximate year the assignments were made, it

does not give an accurate indication as to the amount of spectrum actually built-out by mobile wireless operators. Build-

outs typically lag spectrum assignment dates by a number of years (Rysavy, 2012). There are several reasons. First, some of

the spectrum bands were not cleared of their previous occupants as of the date of their reassignment to mobile wireless.

This was particularly true of the EBS/BRS and AWS-1 bands. Second, build-outs, especially of new technologies,

take planning and time. Thus, for the purpose of evaluating the actual amount of spectrum available to meet customer

demand, the analysis shall employ a rough rule that build-outs of spectrum in new frequency bands do not start until a year

or more following its assignment, and that its full deployment into the market then takes place over an interval of four

6 Not all traffic requires immediate handling. If data bytes supporting a voice conversation are not transferred immediately, the conversation will

become broken up and unintelligible. But data bytes supporting certain file transfers may tolerate delays quite well. Indeed, the speed at which a file istransferred is best indexed by how soon its last byte is transferred, and not by how quickly any of its intermediate bytes arrive. Thus, if file transfer bytes are

held briefly and interspersed into gaps in a voice transmission, it is quite possible for a joint-use wireless channel to operate far more efficiently than if

separate channels were used to satisfy each demand. SeeYuksel et al. (2010).7 Reuse techniques may also be slow to implement as regulatory approvals may need to be obtained before new cell sites are deployed or additional

equipment is placed at pre-existing sites.8 Similar global figures are shown inEricsson (2012a,2012b).9 See Credit Suisse (2012, p. 65) and Ericsson (2012b). Although smartphone penetration may be close to saturation, customers are adding still

additional mobile devices like tablets to their portfolio. Further, widescale adoption of machine-to-machine devices is just beginning ( OECD, 2012). As the

Internet of Things evolves, there may be no limits to how many new data devices will seek to be connected to mobile networks.10 Cisco (2013)projects figures to 2017. To continue the projection out to 2022, Cisco's projected 201617 growth rate of 42% is reduced by 10% each

year to reach a growth rate of 25% in 20212022. This latter growth rate matches reasonably current growth rates in wireline broadband use per subscriber

(Cisco, 2012b).11 The presented traffic growth assumptions may be conservative. Qualcomm believes that mobile data growth from 20102020 will be a factor of

1000 . This figure is much higher than Cisco's assumption of a 103 data growth factor over this period. See http://www.qualcomm.com/media/blog/

2012/08/09/heard-1000x-challenge-hint-it-s-about-mobile-data-growth. Note too, that Cisco's mobile traffic projections include only traffic passing over

mobile carrier RANs. Traffic that mobile devices transmit over Wi-Fi connections, while large and growing, is considered by Cisco to be fixed traffic andincluded its VNI fixed traffic estimates (Cisco, 2013).


R.N. Clarke / Telecommunications Policy ] (]]]]) ]]]]]] 3
http://www.qualcomm.com/media/blog/2012/08/09/heard-1000x-challenge-hint-it-s-about-mobile-data-growthhttp://www.qualcomm.com/media/blog/2012/08/09/heard-1000x-challenge-hint-it-s-about-mobile-data-growthhttp://dx.doi.org/10.1016/j.telpol.2013.11.006http://dx.doi.org/10.1016/j.telpol.2013.11.006http://dx.doi.org/10.1016/j.telpol.2013.11.006http://dx.doi.org/10.1016/j.telpol.2013.11.006http://dx.doi.org/10.1016/j.telpol.2013.11.006http://www.qualcomm.com/media/blog/2012/08/09/heard-1000x-challenge-hint-it-s-about-mobile-data-growthhttp://www.qualcomm.com/media/blog/2012/08/09/heard-1000x-challenge-hint-it-s-about-mobile-data-growth


4/16

years.12 Based on these assumptions it is possible to better compare growth in mobile traffic versus spectrum resources. This is

presented inFig. 3. It is important to emphasize the difference in the scales used on this chart to measure spectrum growth

versus traffic growth. The former is on the left, the latter on the right. Thus, if one looks at 1994 as a base, since that time

1

10

100

1,000

10,000

100,000

1983

1984

1985

1986

1987

1988

1989

1990

1991

1992

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

2010

2011

2012

TB/mon

th-Logarithmicscale

Voice TB Data TB

(1 TB = 1024^4 Bytes)

Fig. 1. History of voice and data demand growth. Notes: Voice TBs calculated fromCTIA (2013) data on mobile minutes, adjusted to data equivalents

assuming 20 kbps encoding for digital networks (2G) and 60 kbps for analog networks (1G). Data TB from Cisco (20092011,2012a and 2013)for 2008

2012, and extrapolated backwards for 20012007.

Fig. 2. Forecasted growth in mobile demand. Notes: Voice and Data TB figures for 20102012 are fromFig. 1. Forecasted 20132022 Voice demand growth

assumes Voice TB remain flat at 2012 levels. Cisco (2013)figures for Data TB are used for 20132017 Data demand. These Data TB figures are extrapolated

forward for 20182022 assuming that Cisco's forecasted demand growth rate for 201617 decays by 10% each year over the 20182022 period.

Table 1

Timeline of U.S. mobile spectrum allocations.

Year Allocation (MHz) Band (MHz) Name

1983 40 850 Cellular

1989 10 850 Cellular

1993 14 800 SMR

1995 130 1900 PCS

2005 194a 2500 EBS/BRS

2006 90 1700/2100 AWS-1

2008 70 700 700 MHz

Total 5 48 (assuming all EBS/BRS spectrum to be usable)

409.5 (assuming only 55.5 MHz of EBS/BRS spectrum)

Notes: Data in this table are assembled fromBazinet and Rollins (2011),FCC (2011, 270275)andFeldman et al.

(2011). Different data sources may suggest slightly different quantities for U.S. mobile wireless spectrum and

allocation dates. These small quantity or timing differences are inconsequential for the purposes of the following

analysis.a Most EBS/BRS spectrum is leased by Clearwire from its primary licensees, who are educational or religious

institutions and providers of wireless CATV services. Because current FCC rules require certain portions of this

spectrum to remain in educational use and may involve other encumbrances, the FCC currently considers only

55.5 MHz of this spectrum to be immediately deployable for mobile wireless use. See http://wireless.fcc.gov/

services/index.htm?job=service_home&id=ebs_brs; http://wireless.fcc.gov/services/index.htm?job=licensing_1&

id=ebs_brs and http://wirelessspectrumreview.com/wireless-spectrum-bands/brsebs/.

12 The exceptions made to this general rule are as follows. Buildouts of the initial Cellular allocation are assumed to have begun in the year of

assignment. For the extra 10 MHz added to the Cellular band in 1989, deployment is assumed to have begun in the same year as assignment and completed

the following year. For the AWS-1 and 700 MHz spectrum bands, it is assumed that buildouts did not commence until two years after the year ofassignment. Note that these assumptions are quite rough, but even moderate changes to them will not alter general patterns in any significant sense.


http://www.qualcomm.com/media/blog/2012/08/09/heard-1000x-challenge-hint-it-s-about-mobile-data-growthhttp://www.qualcomm.com/media/blog/2012/08/09/heard-1000x-challenge-hint-it-s-about-mobile-data-growthhttp://www.qualcomm.com/media/blog/2012/08/09/heard-1000x-challenge-hint-it-s-about-mobile-data-growthhttp://www.qualcomm.com/media/blog/2012/08/09/heard-1000x-challenge-hint-it-s-about-mobile-data-growthhttp://www.qualcomm.com/media/blog/2012/08/09/heard-1000x-challenge-hint-it-s-about-mobile-data-growthhttp://www.qualcomm.com/media/blog/2012/08/09/heard-1000x-challenge-hint-it-s-about-mobile-data-growthhttp://www.qualcomm.com/media/blog/2012/08/09/heard-1000x-challenge-hint-it-s-about-mobile-data-growthhttp://www.qualcomm.com/media/blog/2012/08/09/heard-1000x-challenge-hint-it-s-about-mobile-data-growthhttp://www.qualcomm.com/media/blog/2012/08/09/heard-1000x-challenge-hint-it-s-about-mobile-data-growthhttp://hraunfoss.fcc.gov/edocs_public/attachmatch/DA-11-133A1.pdfhttp://hraunfoss.fcc.gov/edocs_public/attachmatch/DA-11-133A1.pdfhttp://hraunfoss.fcc.gov/edocs_public/attachmatch/DA-11-133A1.pdfhttp://hraunfoss.fcc.gov/edocs_public/attachmatch/DA-11-133A1.pdfhttp://hraunfoss.fcc.gov/edocs_public/attachmatch/DA-11-133A1.pdfhttp://hraunfoss.fcc.gov/edocs_public/attachmatch/DA-11-133A1.pdfhttp://hraunfoss.fcc.gov/edocs_public/attachmatch/DA-11-133A1.pdfhttp://hraunfoss.fcc.gov/edocs_public/attachmatch/DA-11-133A1.pdfhttp://hraunfoss.fcc.gov/edocs_public/attachmatch/DOC-312479A1.pdfhttp://hraunfoss.fcc.gov/edocs_public/attachmatch/DOC-312479A1.pdfhttp://hraunfoss.fcc.gov/edocs_public/attachmatch/DOC-312479A1.pdfhttp://dx.doi.org/10.1016/j.telpol.2013.11.006http://dx.doi.org/10.1016/j.telpol.2013.11.006http://dx.doi.org/10.1016/j.telpol.2013.11.006http://dx.doi.org/10.1016/j.telpol.2013.11.006http://dx.doi.org/10.1016/j.telpol.2013.11.006http://hraunfoss.fcc.gov/edocs_public/attachmatch/DOC-312479A1.pdfhttp://hraunfoss.fcc.gov/edocs_public/attachmatch/DA-11-133A1.pdfhttp://hraunfoss.fcc.gov/edocs_public/attachmatch/DA-11-133A1.pdfhttp://www.qualcomm.com/media/blog/2012/08/09/heard-1000x-challenge-hint-it-s-about-mobile-data-growthhttp://www.qualcomm.com/media/blog/2012/08/09/heard-1000x-challenge-hint-it-s-about-mobile-data-growthhttp://www.qualcomm.com/media/blog/2012/08/09/heard-1000x-challenge-hint-it-s-about-mobile-data-growth


5/16

allocated spectrum has increased from 64 MHz to 548 MHz (assuming all of the EBS/BRS spectrum will be usable) a factor of

8.6. But over the same 19942012 period, total traffic has risen from 900 TB/month to 226,916 TB/month a factor of 256.

To be sure, whileFig. 3provides information about the amounts of spectrum licensed and built out to provide mobile

services, it does not inform directly as to the amount of this lit spectrum that was actually being used to carry the mobile

wireless demands that were offered to these networks in the given years. Such an estimate is essential to determine theamount ofheadroomthat may exist in built-out U.S. mobile networks and may remain available to accommodate increases

in customer demand. For the purposes of evaluating the ability of U.S. mobile networks to accommodate increasing future

demand, it is most useful to have a relatively recent estimate for this headroom. For several reasons, 2010 is chosen. Not only

is 2010 recent, but it is a year that just precedes significant rollouts of services provided over AWS-1 or 700 MHz spectrum.

Further, it is relatively soon after Clearwire began deploying small quantities of service over its EBS/BRS spectrum. Thus, in

2010 it is assumed that the amount of spectrum that was actually being consumed to offer mobile services was 194 MHz

(50 MHz Cellular14 MHz SMR130 MHz PCS).13 But to be especially conservative in estimating needs for additional

spectrum, an additional assumption is made that the 194 MHz of spectrum consumed in 2010 was not used as intensively

(i.e., with as high a fill factor) as may be the case in the future. Based on analyst reports that lit U.S. mobile networks are

operating at only 80% of their capacity on average, the figure this analysis employs for 2010 baseline spectrum usage is

adjusted down to 155.2 MHz (0.80194 MHz).14

From 2009 through 2011 there were no significant additional allocations of mobile wireless spectrum by U.S. regulatory

authorities.

15

But this began to change in 2012 as 12 MHz of unpaired lower 700 MHz spectrum that was previously used forbroadcast fixed or mobile TV was transferred to mobile wireless;16 and further FCC rule changes enabled 20 MHz of

2300 MHz WCS spectrum and 40 MHz of 2200 MHz S-Band MSS spectrum to be used for terrestrial mobile services. 17 As a

result, 72 MHz of additional spectrum may be forthcoming for U.S. mobile wireless and several other bands are also under

study by the FCC and other government authorities for conversion to possible mobile wireless use.18 While these

reallocations or potential reallocations are promising, the effective capacity augmentations they may offer remain

speculative as successful repurposing of these spectrum bands for two-way terrestrial mobile wireless is by no means

assured. Technical or regulatory roadblocks may occur. Thus for future analysis, the analysis assumes, alternatively, that: (a)

no additional spectrum to the 2011 figure of 548 MHz is able to be built out in time to serve customer demand over the

0

50,000

100,000

150,000

200,000

0

100

200

300

400

500

600

1983

1984

1985

1986

1987

1988

1989

1990

1991

1992

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

2010

2011

2012

TrafficinTB/month

MH

zofSpectrum

Spectrum in 4G LTE use

Spectrum in ~3G use

Spectrum in 2G use

Spectrum in 1G use

Total Allocated Spectrum

Traffic

Fig. 3. Spectrum growth versus traffic growth. Notes: Gross spectrum quantities based on Table 1, with author's estimates from note 15 to divide by

technology deployed and date of build-out. Traffic data are fromFig. 1.

13 While it was likely that in 2010 some service was being provided over EBS/BRS and AWS-1 spectrum, these quantities were quite small. Further, it is

also likely that there was still some modest amount of headroom remaining within Cellular, SMR and PCS spectrum as of that date. The effects of these two

over-simplifications are assumed to roughly cancel each other. This assumption of 194 MHz occupancy in 2010 is also somewhat consistent with theestimate of 170 MHz occupancy in 2009 assumed inFCC (2010b)and the assumption of 192 MHz use in 2011 by Bazinet and Rollins (2011).

14 SeeCredit Suisse (2011)for survey results that North American mobile networks are operating at 80% of capacity. Reducing this baseline figure for

effectively utilized spectrum implies that a greater amount of past FCC spectrum allocations remain available to serve future demand thus reducing

implied future needs for additional spectrum.15 Certain attempts were made, however. In January 2011, the FCC issued an Order permitting LightSquared to repurpose 66 MHz of L-Band Mobile

Satellite Service (MSS) spectrum to terrestrial mobile use. See http://hraunfoss.fcc.gov/edocs_public/attachmatch/DA-11-133A1.pdf. This spectrum,

however, was adjacent to spectrum used by Global Positioning System (GPS) satellites and use by LightSquared of this spectrum was ultimately adjudged

to interfere with neighboring GPS services. As a result, in February 2012 the FCC withdrew LightSquared's conditional authorization to use this spectrum

for terrestrial mobile wireless. See http://hraunfoss.fcc.gov/edocs_public/attachmatch/DOC-312479A1.pdf.16 See FCC Order approving the transfer of unpaired 700 MHz spectrum from Qualcomm to AT&Tathttp://hraunfoss.fcc.gov/edocs_public/attachmatch/

FCC-11-188A1.pdf.17 See FCC Orders approving terrestrial mobile use of the WCS band at http://hraunfoss.fcc.gov/edocs_public/attachmatch/FCC-12-130A1.pdf and

approving terrestrial mobile use of the MSS S-band at http://hraunfoss.fcc.gov/edocs_public/attachmatch/FCC-12-151A1.pdf.18 These include up to 120 MHz in the 600 MHz UHF band currently used by broadcast television service and a portion of the 95 MHz in the 1750

1850 MHz band currently used by government agencies. More immediately, it is possible that 20 MHz of government spectrum in the 1755 MHz band

could be paired with 20 MHz in the 2155 MHz AWS-3 band. For a fuller discussion of these future possibilities, see http://transition.fcc.gov/statelocal/presentations/Incentive-Auctions-MB.ppt,NTIA (2012), and more generallyFCC (2010a, pp. 8589).


http://hraunfoss.fcc.gov/edocs_public/attachmatch/DA-11-133A1.pdfhttp://hraunfoss.fcc.gov/edocs_public/attachmatch/DOC-312479A1.pdfhttp://hraunfoss.fcc.gov/edocs_public/attachmatch/FCC-11-188A1.pdfhttp://hraunfoss.fcc.gov/edocs_public/attachmatch/FCC-11-188A1.pdfhttp://hraunfoss.fcc.gov/edocs_public/attachmatch/FCC-12-130A1.pdfhttp://hraunfoss.fcc.gov/edocs_public/attachmatch/FCC-12-151A1.pdfhttp://hraunfoss.fcc.gov/edocs_public/attachmatch/FCC-12-151A1.pdfhttp://transition.fcc.gov/statelocal/presentations/Incentive-Auctions-MB.ppthttp://transition.fcc.gov/statelocal/presentations/Incentive-Auctions-MB.ppthttp://dx.doi.org/10.1016/j.telpol.2013.11.006http://dx.doi.org/10.1016/j.telpol.2013.11.006http://dx.doi.org/10.1016/j.telpol.2013.11.006http://dx.doi.org/10.1016/j.telpol.2013.11.006http://dx.doi.org/10.1016/j.telpol.2013.11.006http://transition.fcc.gov/statelocal/presentations/Incentive-Auctions-MB.ppthttp://transition.fcc.gov/statelocal/presentations/Incentive-Auctions-MB.ppthttp://hraunfoss.fcc.gov/edocs_public/attachmatch/FCC-12-151A1.pdfhttp://hraunfoss.fcc.gov/edocs_public/attachmatch/FCC-12-130A1.pdfhttp://hraunfoss.fcc.gov/edocs_public/attachmatch/FCC-11-188A1.pdfhttp://hraunfoss.fcc.gov/edocs_public/attachmatch/FCC-11-188A1.pdfhttp://hraunfoss.fcc.gov/edocs_public/attachmatch/DOC-312479A1.pdfhttp://hraunfoss.fcc.gov/edocs_public/attachmatch/DA-11-133A1.pdf


6/16

20122022 period;19 and (b) the FCC is successful in reallocating a total of 300 additional MHz of spectrum to mobile

wireless and this spectrum is built out over the 20142022 period.20 The former may be considered a pessimistic

assumption and the latter an optimistic one.21 It is likely that actual effective spectrum additions will range somewhere

between these two figures.22

3.3. Intensity of spectrum reuse

Fig. 3shows that mobile usage has grown orders of magnitude faster than available spectrum, thus provides telling

evidence that so far the largest contributor to increased mobile capacity in the U.S. has been investment by mobile operators

to evolve networks to more spectrally-efficient technologies and to erect more cell towers that enable more intensive reuse

of the radio frequencies that have been allocated to mobile wireless. When one examines the data, it is clear that most of

this capacity uplift has been due to more intensive reuse. From 1985 to 2012, cell site counts have grown by a factor of 330.

This history of cell site expansion is displayed inFig. 4.

While this may suggest that capacity reuse has also risen by a factor of 330, the capacity effects of this proliferation of cell

sites are hard to quantify precisely. For example, in the early years of a mobile technology's deployment, it is likely that most

new sites were deployed to extend coverage of that technology, and not simply to increase the capacity of that technology

within already covered areas. In the later years of a technology's deployment, capacity increases are more likely to be the

purpose of new sites. But, because later deployments may be disproportionately of micro or pico cells, which are designed to

serve smaller areas and traffic quantities, incremental capacity expansion per-site may decrease as cell site counts rise.23

Butif it is generously assumed that all recent site deployments have been strictly to improve capacity, and that these

deployments provide an effective uplift in network capacity in direct proportion to their number, this suggests that if site

deployment rates observed over the last six to ten years continue, traditional reuse will provide about an 8% per year

increase in capacity to serve demand growth.

4. Spectral efciency and unfolding 4G LTE technological capabilities

4.1. Evolution of U.S. spectral efficiency and reuse

Over the past thirty years there have been two sources of improved spectral efficiency of U.S. mobile networks:

deployments of wireless technologies that are more spectrally efficient; and increasing the quantity of spectrum used by

-

50,000

100,000

150,000

200,000

250,000

300,000

1985 1988 1991 1994 1997 2000 2003 2006 2009 2012

Fig. 4. Number of U.S. cell sites. Notes: CTIA (2013)count of cell sites. Note that CTIA's count is by network technology by physical location.

19 Note that while this assumption may appear to neglect the contribution of the 72 MHz that has been newly provided in 2012, it also presumes that

all 194 MHz of EBS/BRS spectrum previously assigned to mobile wireless (most of which still remains fallow) will be deployed over the 2012 2022 period.20 The FCC's National Broadband Plan (issued in March 2010) calls for an additional 300 MHz of spectrum to be allocated to mobile wireless by 2015,

and for a total of 500 MHz to be added by 2020. SeeFCC (2010a, p. 84).21 Realistically, given that only 72 MHz has been added since the date of the National Broadband Plan, without significant changes in U.S. spectrum

regulatory policy there is very little probability that a further 228 MHz of spectrum could be provided and built out within the 2014 2022 timeframe to

meet even the National Broadband Plan's2015 target of 300 MHz in additional spectrum. Lead times for identifying, auctioning, clearing and building out

spectrum are generally on the order of 810 years. SeeFCC (2010a, p. 79).22 A recent U.S. government report (PCAST, 2012) recommends that 1000 MHz of government-controlled spectrum within the sub-4000 MHz bands be

made available for shared, secondary use by commercial mobile wireless. While ambitious, it is extremely unclear as to whether this proposal is practicable

from a technical or economic point of view, and even if it is, whether it can have any appreciable effect within the 20142022 time period. As noted earlier,

North American mobile networks already operate at 80% utilization within their exclusively licensed bands. Thus, the opportunity to share, on an

idiosyncratic secondary basis, some additional spectrum, poses substantial economic challenges. New network equipment and cell sites would need to be

deployed, only to be able to make partial unassured use of these new bands. This suggests that the additional costs of using these shared bands may be

disproportionate to their effective capacity uplift. Further, it is unlikely that all frequencies would be available throughout the country or in the particular

locations most in need of additional capacity.23 SeeChapin and Lehr (2011, p. 32),FCC (2010a, p. 77),Goldstein (2012)and Paolini (2011).




7/16

these more efficient technologies relative to the quantity of spectrum used by older, less efficient technologies. For thepurposes of the analysis presented here, 1G technologies are assumed to have a spectral efficiency of less than 0.1 bps/Hz on

a sector basis; 2G technologies have an efficiency of 0.25 bps/Hz; 3G technologies an initial efficiency of 0.5 bps/Hz

rising to between 0.90 bps/Hz and 1.0 bps/Hz for advanced implementations such as EV-DO, Rev. A or 4G HSPA ;24 and

current 4G LTE technologies 1.4 bps/Hz.25 Based on these spectral efficiency figures and estimates of the quantities of U.S.

mobile spectrum deployed to different technologies, it is possible to map the evolution of the potential capacity of U.S.

mobile networks to handle voice and data traffic. This is displayed by the Spectral Capacity Index in Fig. 5(scale on right side

of the chart). As can be seen, the deployment of more spectrum and the implementation of more efficient 3G and 4G

technologies have provided a substantial uplift in mobile network carrying capacity.

But in addition to capacity growth due to spectrum expansion and the migration from older technologies to newer ones,

there has also been increased reuse of frequencies due to the deployment of additional cell sites (Reuse Index). These two

capacity expansion factors can be combined to estimate the total growth in U.S. mobile network capacity (Total Capacity

Index).26 This index is displayed inFig. 6where it is compared against an index of past traffic growth.

While mobile network capacity expansion has been impressive in its ability to roughly match past rates of demand

growth, mobile capacity expansion faces an even more daunting task in keeping up with the projections for future demand

growth displayed inFig. 2. Whether or not it will be successful will depend importantly on future contributions to capacity

growth from advances in 4G LTE technology. These will be discussed in the following section.

4.2. Efficiency improvements from future 4G LTE innovations

While current LTE networks are significantly more spectrum-efficient in carrying mobile traffic than earlier network

technologies, LTE is capable of even further improvements as its technology progresses to LTE-Advanced (LTE-A or 4G).27

These improvements divide into three categories: increasing the raw transmission throughputs over LTE radio links; further

increasing the possibilities for spectrum reuse; and packing offered traffic more efficiently into available transmission

capacity.28

0

10

20

30

40

50

60

70

80

90

100

0

50

100

150

200

250

300

350

1983

1984

1985

1986

1987

1988

1989

1990

1991

1992

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

2010

2011

2012

RelativeSpectralCapacity(1985=1)

MHzofS

pectrum

Spectrum in 4G LTE use

Spectrum in ~3G use

Spectrum in 2G use

Spectrum in 1G use

Spectral Capacity Index

Fig. 5. Capacity expansion from spectrum expansion and migration of users to more spectrally-efficient technologies. Notes: Division of spectrum by

technology and dates of build-out are fromFig. 3. Spectral efficiency factors used to compute spectral capacity are fromRysavy Research (2012b, p. 55).

24 EV-DO stands for Evolution-Data Optimized and HSPA stands for High Speed Packet Access. Although HSPA technology has very high speed

performance (21 Mbps and above) and is therefore recognized as a 4G technology by the International Telecommunications Union (ITU, 2010); its spectral

efficiency is similar to that of 3G technologies. Therefore, for the purposes of the spectral efficiency analysis (not speed analysis) used in this paper, 4G

HSPA is included in the 3G category.25 SeeRysavy Research (2012b, pp. 5455)and Rysavy Research (2011b). These figures represent the consensus view of 4G Americas contributors of

the actual mobile spectral efficiencies expected under real-world network conditions. As such, they are lower than other estimates developed under

simulated ideal network conditions. While these efficiency figures may be imprecise, all that is absolutely necessary is that they provide a reasonably

accurate picture of the relative spectral efficiencies of the presented technologies.26 Note that the presented capacity growth index is illustrative only. This is because this index does not account for possible changes in network

packing over the 19852012 period. SeeSection 4.2.3for a further discussion of network packing as a possible source of effective capacity growth.27 The progression of LTE technology is indexed by different releases of 3GPP specifications with initial LTE-A specifications in Release 10. See Rysavy

Research (2012b, pp. 2426).It is possible that certain of the enhancements termed 4Gare part of Releases earlier than 10, but since they are only now

in the process of first implementation in networks, they will, for convenience, be called 4G .

28 Lawson (2012)andReal Wireless (2010)also provide catalogs of capacity expansion techniques. See Bhat et al. (2012)for more technical discussionsas to how LTE-A achieves these improvements.




8/16

4.2.1. Improved transmission throughputs

Perhaps the most dramatic potential improvements in LTE's capacity to handle traffic derive from higher-order Multiple

Input Multiple Output (MIMO) implementations.29 Current LTE deployments use 22 MIMO. This places two antennas at

the base station and two antennas in the user device.30 Because of the slight physical displacement of each transmitting

antenna from the other transmitting antenna(s) and of each receiving antenna from the other receiving antenna(s), eachsent and received signal will be subject to different multipath characteristics.31 By examining the four signals together, more

of the originally encoded information may be extracted. In addition, MIMO technology may be used to send multiple

concurrent transmission streams between the base station and user device. This too may increase throughput over a given

amount of spectrum32 Implementations of 42, 4 4, and even 88 MIMO are already part of LTE-A specifications. While

theoretically, 44 MIMO should provide twice the throughput of 22 (although at higher power usage), and 88 MIMO

four times the throughput of 22, field tests suggest that actual uplifts are substantially less (Rysavy Research, 2012b, p. 56).

However, these important advances come with their own challenges. First, mobile user devices generally are quite small.

Although it has been possible to place two diverse antennas within their form factor, placing four or eight diverse antennas is

much more problematic. Indeed, it may be that throughput advances from higher-order MIMO will only be won at the cost of

physically larger and more power-hungry mobile devices. Second, placing additional radio transceivers at cell sites also bears

increased costs for equipment, tower structure and electric power.

Another technology being developed is called Coordinated Multi-Point transmission/reception (CoMP). This technology

leverages what would otherwise be undesirable interfering transmissions from adjacent cells or sector antennas by havinguser devices purposefully communicate with several cell sites (or sector antennas at a single cell site) simultaneously.

Because CoMP coordinates these transmissions, what was previously interference from adjacent cell sites or sector antennas

is actually converted into useful information-carrying bandwidth, and a more robust total signal received (Rysavy Research,

2012b, pp. 103105). While potentially valuable, CoMP is expected only to enter LTE-A specifications via 3GPP Release 11 in

late 2012, and its potential implementation in commercial networks is at least several years in the future.

However, these two significant throughput-improving advances associated with LTE-Advanced are not achieved just

through software tweaks. Higher-order MIMO requires new antennas at base stations, and completely new user devices that

incorporate both the increased number of antennas and the chipsets necessary for the more complex processing of received

and sent signals. New chipsets will also be needed for advanced versions of CoMP. As a result, it is unlikely that commercial

networks fully incorporating these advances will be seen until 2016 or beyond.

Taking all of these potential enhancements together, LTE-A technology is expected to achieve a spectral efficiency of

2.25 bps/Hz a lift of nearly 61% over the 1.4 bps/Hz efficiency of current 2 2 MIMO LTE deployments (Rysavy Research,

2012b, p. 55).

1

10

100

1,000

10,000

100,000

1985

1986

1987

1988

1989

1990

1991

1992

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

2010

2011

2012

Logari

thmicscale(1985=1)

Reuse Index

Spectral Capacity Index

Total Capacity Index

Traffic Index

Fig. 6. Contributions of increasing spectral efficiency and frequency reuse to overall mobile network capacity. Notes: Reuse figures fromFig. 4, spectral

capacity figures fromFig. 5and traffic figures from Fig. 1. Total capacity index is the product of these two component indices.

29 MIMO technology employs multiple antennas on cell towers and within user devices to enable two or more parallel radio links connecting a tower

to a user device. See Korowajczuk (2011, Chapter 14)and Rysavy Research (2012b, pp. 74, 102103).30 The first figure in AB MIMO represents the number of antennas at the network base station. The second figure represents the antenna count at the

user device. 22 MIMO results in four effective signals: one from transmitting antenna 1 to receiving antenna 1; a second from transmitting antenna 1 to

receiving antenna 2; a third from transmitting antenna 2 to receiving antenna 1; and a fourth from transmitting antenna 2 to receiving antenna 2.31 Multipath interference is perhaps the greatest reason why mobile technologies fail to attain throughput rates that match their theoretical maxima.

This interference results from environmental obstacles like trees, buildings or mountains reflecting radio beams so that all of a signal's energy does not take

the same path from its source antenna to its destination antenna. Because the whole signal does not take the same path (multipath), portions of the signal

arrive at the destination both faded and slightly delayed, creating a less intelligible received signal. This becomes most acute when the signal is of high

bandwidth because the portion of a data symbol whose receipt is delayed by multipath may interfere with the portion of a subsequent data symbol that

arrives on time.

32 There are even more reasons why MIMO technology may improve throughputs between two points. Their discussion is beyond the scope of thispaper. For further information, see Korowajczuk (2011, pp. 262281).




9/16


10/16

reuse of available radio frequencies will be intensified. Because of the very important capacity-expansion capabilities of this

LTE-A innovation, the analysis assumes that it results in an annual growth rate for effective LTE-A cell sites of 16% 38 a

figure double the historic 8% growth rate for non-LTE-A sites thus implying a highly accelerated growth rate for LTE-A

frequency reuse.

4.2.3. More efficient data packing

Opportunities to pack data more efficiently into total available radio access spectrum are limited in mobile networks.

Perhaps the greatest source of inefficiency in data packing is due to geography. In less populated areas of the country,portions of assigned mobile wireless spectrum may be fallow because local demand is insufficient to exhaust its carrying

capacity. In contrast, in dense urban areas there are concentrated sources of demand that exhaust available mobile network

capacity. While it is possible (and economic) for large users of resources like electricity (e.g., aluminum smelters) to locate

close to remote hydroelectric dams, subscribers using mobile devices want to use mobile services where they are currently

located (which generally is in denser population areas) and not in areas where they are not located.39

One way to gain packing efficiencies is to optimize data content sought by mobile subscribers into lower bandwidth

format (e.g., stripped down web pages or video files downconverted to resolutions more appropriate for smaller format

mobile devices).40 Efficiency can also be gained from providing different mobile data applications with Quality of Service

(QoS) that is tailored more precisely to their individual needs. For example, Voice over Internet Protocol (VoIP) or

videoconferencing needs a consistent low-latency stream of data bytes. Errored packets, while they may create some

degradation, can typically be accommodated by these applications without retransmission. In contrast, web browsing may

function best if intermittent bursts of data packets arrive at high speed. Finally, file transfer bytes may be unforgiving of

packet errors, but accommodate latency quite well. If a mobile network is offered a variety of these different applicationtypes by its subscribers, then by using QoS-based packet scheduling it can accommodate simultaneously these diverse

demands using less total bandwidth than if each application is provided with the same QoS. By incorporating QoS-based

packet scheduling into its specifications, newer releases of LTE will be able to allot bandwidth to different applications in a

more efficient manner than earlier mobile technologies. This will permit operators to offer high quality Voice over LTE

(VoLTE) service and eventually to decommission their separate 2G and 3G voice networks. By enabling increases in the

total amount of service customers can be offered through a limited amount of spectrum, it is possible that LTE QoS and

packet scheduling could improve effective throughput by up to 20%.41

In addition to more efficient data packing enabled by advanced technology, it is important to note that ordinary changes

in subscriber use patterns may result in more or less efficient network fills. For example, as mobile data services evolve from

ones with a business focus to consumer focus, it is possible that tomorrow's data will be offered to the network more evenly

throughout the day than in prior years. Thus, if a significant amount of the new video traffic flooding U.S. mobile networks is

off-peak load it may be carried with reduced requirements for increased network capacity. Of course, the opposite may

occur, too.

5. Overall ability of technology and spectrum to meet forecasted demand

To measure the ability of these various expected enhancements to mobile network capacity to meet forecasted growth in

demand, the analysis employs a model similar to that used in FCC (2010b). This model is assumes that if U.S. mobile

networks have successfully used a particular set of technologies, number of cell sites, and amount of spectrum in a recent

base yearto serve that year's level of demand, then it is possible to use these base year ratios to project the capability of

future networks incorporating additional amounts of spectrum, an evolved set of technologies and an expanded number of

cell sites to serve forecasted demand.42 To accomplish this, the analysis compares an index of forecast demand growth

against an estimated index of overall capacity expansion. This latter capacity index is itself the multiplicative product of

separate estimated indices for: (a) growth in raw spectrum resources; (b) customer migration to more spectrally-efficient

higher-G mobile technologies; (c) growth in effective spectrum reuse; and (d) additional capacity improvement due to moreefficient network packing enabled by LTE and LTE-A technologies.

38EffectiveLTE-A sites are sites that are equivalent to traditional macro or micro cells in their capacity uplift. But note that because each small cell is

expected to handle less traffic than a large cell, this implies that actual small cell counts will be substantially larger than the effective cell counts that are

estimated here.39 Of course the classic exception to this rule is when a mobile customer is in an area of poor reception and moves a short distance (say, to a higher

floor of his house or out of a sheltered building alcove) to get a better signal.40 Transcoding video traffic and limiting streaming buffer sizes is an optimization technique already employed by certain mobile operators. See Verizon

Wireless: Data Plans & Features Terms and Conditions, retrieved from: http://support.verizonwireless.com/terms/products/vz_email.html and

Optimization Deployment Terms & Conditions, retrieved from:http://support.verizonwireless.com/terms/network_optimization.html.41 SeeYuksel et al. (2010)for a more in-depth explanation for why multi-use networks incorporating QoS require less bandwidth for equivalent service

quality than multiple single-purpose networks. LTE-A carrier aggregation may be an additional source of packing efficiency. By permitting non-contiguous

spectrum blocks to be aggregated into wider channels, and by permitting asymmetric uplink and downlink block sizes, it will be possible for mobile

operators to increase the throughput capacity of their spectrum holdings. This uplift is also considered in the proposed 20%.

42 Note that this model abstracts from a location- or time-specific analysis by assuming that future demand will resemble base year demand in termsof its location and temporal profile. To the extent that this profile of demand characteristics changes over time, the model will be less precise.


http://support.verizonwireless.com/terms/products/vz_email.htmlhttp://support.verizonwireless.com/terms/network_optimization.htmlhttp://support.verizonwireless.com/terms/network_optimization.htmlhttp://dx.doi.org/10.1016/j.telpol.2013.11.006http://dx.doi.org/10.1016/j.telpol.2013.11.006http://dx.doi.org/10.1016/j.telpol.2013.11.006http://dx.doi.org/10.1016/j.telpol.2013.11.006http://dx.doi.org/10.1016/j.telpol.2013.11.006http://support.verizonwireless.com/terms/network_optimization.htmlhttp://support.verizonwireless.com/terms/products/vz_email.html


11/16

The first capacity index component addressed is raw spectrum growth. Two alternative indices for growth of raw

spectrum are developed, one as a baseline and another as an upper bound. The baseline is an index of built-out spectrum

that assumes that no more spectrum allocations are forthcoming to the U.S. mobile wireless industry beyond the 548 MHz

already allocated and that all of this spectrum is usable. The upper bound index assumes that an additional 300 MHz of

usable spectrum will become available to the industry and built-out over the 20142022 time period. The year 2010 is

assumed to be the model's base year. Although the analysis estimates that networks employing 294.5 MHz of allocated

spectrum were built-out as of that year, it assumes that 2010 demand required only 155.2 MHz of this spectrum to be filled.

This implies an initial Spectrum Index value of 1.90 ( 294.5 MHz/155.2 MHz) to reflect the headroom available in 2010

spectrum deployments.43 The baseline index is shown in column (c) ofTable 2. Column (d) shows the assumed timeline for

building out the additional 300 MHz that may be offered to the industry.44 An upper bound Spectrum Index incorporating

these additional allocations is given in column (f).The next component of the capacity index developed are two alternative indices for effective spectral efficiency. They are

presented in Table 3 and constructed by weighting the relative spectral efficiencies of each available mobile wireless

technology by estimates of the fraction of total built-out spectrum that will be in use by these different technologies in each

year. The baseline index assumes that only already-allocated spectrum will be available over the 20142022 period. The upper

bound assumes an additional 300 MHz of spectrum will be released to the industry and that nearly all of it will be built out as

LTE-A thus raising the industry's average spectral efficiency. Column (g) ofTable 3shows the baseline index for capacity

growth resulting from migration of use to higher G technologies assuming current total spectrum allocations. 45 Column (m)

shows the upper bound index assuming spectrum allocations are augmented by 300 MHz over the 20142022 period.

Table 4displays forecast growth in spectrum reuse. Column (b) shows forecasted cell site counts assuming continued 8%

annual growth and current raw spectrum allocations. The index associated with this is in column (c). But as noted, LTE-A

(4G) enables Het-net development, thus facilitating accelerated growth of small LTE-A cells. This acceleration is assumed

to double the annual growth of effective LTE-A cells to 16%. This is shown in columns (d), (e) and (f). If an additional

300 MHz of spectrum is supplied, this changes upwards the fraction of cell sites that will be deployed as LTE-A. Adjustedreuse growth figures to represent this are provided in columns (g), (h) and (i).

The final source of increased capacity is more efficient network packing. As noted, QoS-based packet scheduling, carrier

aggregation and VoLTE implemented in LTE are assumed to permit a phased-in 20% improvement in LTE and LTE-A

throughputs. This is shown inTable 5 assuming LTE and LTE-A service prevalence both with current spectrum allocations

(columns b and c) and with 300 MHz of additional spectrum (columns d and e).

Fig. 7shows the contributions of each of the above-developed sources of capacity growth with solid lines assuming the

baseline of no additional spectrum and dashed lines assuming an additional 300 MHz of spectrum. As can be seen,

Table 2

Contributions of raw spectrum to mobile capacity growth.

(a) (b) (c) (d) (e) (f)

Year Currently allocated spectrum

(built-out MHz)

Baseline

Spectrum Index

Additional built-out

allocations (MHz)

Augmented spectrum

quantity (MHz)

Upper bound

Spectrum Index

2010 294.5 1.90

2011 334.5 2.16

2012 374.5 2.412013 426.6 2.75 2.75

2014 478.8 3.08 30 508.8 3.28

2015 513.4 3.31 30 573.4 3.69

2016 548.0 3.53 30 638.0 4.11

2017 548.0 3.53 30 668.0 4.30

2018 548.0 3.53 40 708.0 4.56

2019 548.0 3.53 40 748.0 4.82

2020 548.0 3.53 40 788.0 5.08

2021 548.0 3.53 30 818.0 5.27

2022 548.0 3.53 30 848.0 5.46

Notes: Time pattern for deployment of already allocated spectrum assumes that substantially all of Clearwire's EBS/BRS spectrum is built out by 2016

(see note 13). Deployment pattern of the 300 MHz of additional spectrum assumes that WCS and unpaired 700 MHz spectrum will be built out in 2014 and

PCS H-block and MSS S-band spectrum begin to be deployed in 2015 (see notes 18 and 19). Spectrum deployed from 2016 to 2022 will consist of 600 MHz

UHF spectrum, AWS-3 spectrum and additional MSS spectrum or small cell 3550 MHz spectrum (see notes 20 and 35).

43 SeeSection 3.2, supra.44 SeeRysavy Research (2012a) for a discussion of time paths and technical and business considerations governing mobile network build outs.45 This migration of customers to higher G technologies (principally from 2G to 3G, 4G or 4G), assumes no government regulations restricting U.S.

mobile operators from migrating their customers. In fact, it was not until 2002 that the FCC issued an Order permitting U.S. networks to retire their 1Gnetworks in 2008. See http://hraunfoss.fcc.gov/edocs_public/attachmatch/DOC-225216A1.pdfand Lawson (2008).


http://hraunfoss.fcc.gov/edocs_public/attachmatch/DOC-225216A1.pdfhttp://dx.doi.org/10.1016/j.telpol.2013.11.006http://dx.doi.org/10.1016/j.telpol.2013.11.006http://dx.doi.org/10.1016/j.telpol.2013.11.006http://dx.doi.org/10.1016/j.telpol.2013.11.006http://dx.doi.org/10.1016/j.telpol.2013.11.006http://hraunfoss.fcc.gov/edocs_public/attachmatch/DOC-225216A1.pdf


12/16


13/16

Thus, if the U.S. mobile industry is successful at: (a) deploying all of the raw spectrum assumed; (b) increasing its reuse

intensity of this spectrum; (c) achieving all of the stated LTE and LTE-A efficiency lifts; and (d) quickly migrating customers

to these higher G technologies; these individual contributions may be multiplicatively combined and their joint lift on

mobile capacity compared with forecasted traffic demand. This is done in Fig. 8.The solid red line shows baseline capacitygrowth assuming current spectrum allocations. The dashed red line shows the upper bound estimate of capacity growth

assuming an extra 300 MHz of spectrum. This capacity growth is in comparison to the Cisco VNI-based demand forecast for

U.S. mobile traffic shown in solid blue.

As can be seen, assuming baseline mobile capacity growth, the U.S. goes into capacity deficit in 2014; and by 2020 demand

will be more than double available capacity. While beginning to add 300 MHz of additional spectrum in 2014 narrows somewhat

the capacity gap, even this is inadequate to keep the U.S. out of deficit beyond 2017. Indeed, to keep the U.S. out of deficit through

2022, the modeling suggests that total spectrum additions over the 20142022 period need to amount to roughly 560 MHz for

a total of 1108 (548560) deployed MHz.46 The index figures underlyingFig. 8are provided inTable 6.

Table 5

Contributions of improved network packing to mobile capacity.

(a) (b) (c) (d) (e)

Year With current spectrum With upper bound spectrum

Lift due to 4G and 4G packing efficiency Baseline

Network Packing Index

Lift due to 4G and

4G packing efficiency

Upper

Network Packing Index

2010 1.002011 1.00

2012 1.0% 1.01

2013 2.8% 1.03

2014 5.9% 1.06 6.0% 1.06

2015 10.0% 1.10 10.0% 1.10

2016 11.5% 1.12 11.8% 1.12

2017 12.3% 1.12 12.8% 1.13

2018 13.4% 1.13 14.0% 1.14

2019 14.3% 1.14 15.0% 1.15

2020 15.4% 1.15 16.3% 1.16

2021 16.9% 1.17 17.4% 1.17

2022 18.3% 1.18 18.7% 1.19

Ultimate 4G and 4G packing efficiency lift: 20%

Notes: Improvement in LTE and LTE-A network packing is assumed to be 5% in 2012, 10% in 2013, 15% in 2014 and 20% thereafter. These effects are weightedby the prevalence of 4G and 4G technologies in overall U.S. mobile wireless networks. Assumption for ultimate effect of QoS-based packet scheduling is

author's estimate based on VoLTE implementation, carrier aggregation and Yuksel et al. (2010).

1.00

1.50

2.00

2.50

3.00

3.50

4.00

4.50

5.00

5.50

2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022

GrowthInd

ices(2010=1.0

0)

Raw Spectrum

Migration to Higher G Technologies

Reuse with 4G+ Het-net Lift

Network Packing

Dashed lines assume 300 MHz

spectrum augment

Fig. 7. Contributions to mobile capacity growth. Notes: Solid lines assume baseline of no additional spectrum allocations. Dashed lines show contributions

if 300 MHz of additional spectrum is allocated.

46 This figure, while large, is smaller than certain other estimates of incremental spectrum needs. In 2009, CTIA projected the U.S. to require an

additional 800 MHz of allocated spectrum off of a base of 410 MHz for a total of 1210 MHz (CTIA, 2009). Note, however, CTIA employed more conservativeassumptions about potential increases in spectrum reuse than the aggressive (and possibly unrealizable) assumptions employed in the present model.




14/16

While this analysis suggests that evolving demand for U.S. mobile wireless services is likely to be stymied by inadequate

capacity growth, it is also useful to consider possible reasons why this may fail to occur or if it does occur, how the market

will equilibrate.

One possibility is that the forecasted demand figures are wrong and mobile service usage will not expand at the rates

forecasted by Cisco. While this is certainly a possibility, it should be noted that until 2012, Cisco's updates of its Global

Mobile Data Traffic Forecast generally found actual Total Global Mobile Data Traffic to be larger than what its forecast from

the previous year predicted (Cisco, 20092011,2012a). While Ciscos 2013forecast was a revision down from its earlier

forecasts, it is also this more conservative forecast that is used by the analysis to project future demand.47

Another possibility is that far greater load-shifting, and thus improved network packing, will take place than suspected.To the extent that disproportionate amounts of new customer demand are for mobile services at times-of-day and in

geographical locations where network capacities are not at their limit, it is possible that networks could absorb increased

traffic without requiring proportional capacity reinforcements. Absent very granular network traffic data, it is impossible to

know the extent of this possible mitigating effect.

On the other hand, it is more likely that reductions in service quality or price increases will end up being the principal

forces for equilibrating the market. When wireless capacities are tight, data connections will slow down. Either customers

will accept slower performance of their mobile applications, or they will discontinue their use, or transfer their use to off-

peak periods or locations. It is also possible that they will eschew particularly data-hungry applications in favor of less-

desirable substitute applications that have the virtue of reduced data use. Either way, the effective service quality that

customers receive will be reduced. Further, it is quite certain that prices will also be a major equilibrator of the market.

1

10

100

2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022

GrowthIndices-Log

arithmicscale(2010=1)

Capacity Index with Current Spectrum

Capacity Index with Upper Bound Spectrum

Demand Index

Fig. 8. Comparing mobile capacity growth to demand growth. Notes: Capacity indices developed fromFig. 7and demand indices from Fig. 2.

Table 6

Mobile capacity growth versus demand growth.

(a) (b) (c) (d)

Year Total indices

Capacity Index with current spectrum Capacity Index with upper b ound spectrum Demand Index

2010 1.90 1.00

2011 2.98 1.91

2012 4.40 3.23

2013 6.38 5.24

2014 8.60 9.19 8.47

2015 12.09 13.41 13.08

2016 15.85 18.80 19.22

2017 18.26 23.24 27.13

2018 21.71 29.82 37.26

2019 25.64 37.51 49.82

2020 30.96 47.90 64.97

2021 38.15 60.04 82.79

2022 47.13 76.49 103.25

47 SeeCisco (2013).Also remember that Qualcomm projects 1000 data traffic growth from 20102020 far greater than the 103 growth factorfor data traffic and the 65 growth factor for total traffic implicit in the analysis' Cisco VNI-based figures for 20102020. See note 11.




15/16


16/16

Documents

1. Expanding Mobile Wireless Capacity -s2.0-S0308596113001900-Main